The first aim of the present proposal is to differentiate among several seemingly similar forms of stimulus classes. We approach these stimulus classes with the working hypothesis that they are hierarchical. The assumption is based on the empirical literature in combination with a logical analysis of the skill requirements of the various classes. For example, we expect that classes that require conditional discrimination of visual stimuli and bi-directional transfer may be more difficult than classes that require conditional position discriminations and only unidirectional transfer. Thus, the latter may be more likely in individuals with mental retardation. The second aim is to investigate the possibility of building higher level stimulus classes from lower level ones. We will attempt to demonstrate stimulus classes in the more complex procedures by first demonstrating the same classes in less complex procedures, and then gradually withdrawing this support over exemplars. The studies will help to precisely characterize cognitive deficits of individuals with retardation. Further, they will explore strategies for the remediation of these deficits.